background image
Generating and Using the Knowledge Space
The RDF cards represent the original XML
documents at the semantic level.The union of such
RDF cards constitutes a knowledge base, which is
a harmonised semantic representation of the under-
lying heterogeneous databases.
However, so far the RDF instance descriptions
have not left the museum.The museum has complete
control of the information it wants to publish, and it
does not need to allow the FMS system access to its
internal database system.The RDF data are placed in
a public directory on the museum's WWW server.
The Web crawler of the FMS system harvests the
instance descriptions from the different museums, and
the system combines them into an RDF repository.
This repository is a large semantic graph that consists
of the shared ontology and metadata.
How does a user now search and navigate in this
knowledge space? In the FMS system, this is imple-
mented by a server-side software, called Ontogator.
Based on the semantic graph, this software dynami-
cally generates semantic linkages for the user's Web
browser.
One way of using the FMS system is view-based
filtering.The user can select classes of resources from
the ontology, and the system finds the instances that
match the selected class restrictions. By constraining
classes (views) further, the collection instance data
searched for are eventually found.
The software also supports topic-based navigation
by providing semantic links between topics of inter-
est, the creation of which is based on the collection
domain ontology and the related metadata of the
collection records.This means that the links also
provide the user with an impression of the wider
context and pragmatics of the objects in the
museums' collections.
From human users to software agents
As described in the Finnish Museums on the
Semantic Web example, the RDF repository is a
large semantic graph that consists of the shared
ontology and metadata of the participating museums.
Such a repository can be queried and the results, a set
of pointers to the relevant resources, can be accessed
using Web browsers.The opportunities provided by a
system like the one developed by the FMS initiative
(e.g. topic-based navigation) are at present restricted
primarily to human users.
The Semantic Web vision includes intelligent soft-
ware agents which `understand' semantic relationships
between Web resources and seek relevant information
as well as perform transactions for humans.
14
This software would be capable of autonomous
action, i.e. could run without direct human control
or constant supervision, and ideally is very flexible in
doing this. Characterisations of this flexibility include
actions that are `reactive, `proactive', and `social' (see
below).
While the basic idea of agents is very intuitive and
appealing, the actual theory is complex, the tools are
immature, the solutions small and prototype-based.
In fact, as a parallel distributed systems technology,
agents belong to the most complex class of software
technology.
However, this primer will conclude with a sum-
mary of what an intelligent software agent is and
what such a software would generally be capable of
doing.This should also serve as an indication of how
great the challenge for research and technological
development is to make the full Semantic Web vision
a reality.
Intelligent Software Agents
The following definitions are taken from Michael
Wooldridge's introduction to multiagent systems
15
:
Agent:
`An agent is a computer system capable of autono-
mous action in some environment'.
Intelligent agent:
`An intelligent agent is a computer system capable
of flexible autonomous action in some environment'.
Flexible autonomous action:
`By flexible autonomous action, we mean reactive,
proactive, social.'
| Reactivity: `A reactive system is one that maintains
an ongoing interaction with its environment, and
responds to changes that occur in it (in time for
the response to be useful)'.
| Proactiveness: `An agent serves a purpose, and
therefore exhibits goal-directed behaviour, in-
cluding the capacity to recognise opportunities
for useful courses of action'.
| Social ability: `Social ability in agents is the ability
to interact with other agents (and possibly humans)
via some kind of agent communication language,
and perhaps cooperate with others'.
Desirable further properties of agents are:
| Mobility: the ability to move around an electronic
network;
| Rationality: an agent will act in such a way that it
does not prevent itself from achieving its goals (as
far as this is possible with a limited set of beliefs
representing its world knowledge);
| Learning: an agent will improve its performance
over time.
DigiCULT 35
14
T. Berners-Lee, J. Hendler,
O. Lassila, Scientific American,
May 2001,
http://www.sciam.com/2001/
0501issue/0501berners-lee.html
15
M.Wooldridge:
An Introduction to
Multiagent Systems.
Chichester:Wiley 2002, and
http://www.csc.liv.ac.uk/
~mjw/pubs/imas/